Long‐term fatigue following aneurysmal subarachnoid haemorrhage and the impact on employment

Abstract Background and purpose Fatigue is common following aneurysmal subarachnoid haemorrhage (aSAH) but little is known about its frequency, prognosis and impact on employment. The aim of this study was to assess the frequency of fatigue, whether it changes over time and the relationship to employment in the long term. Methods This was a retrospective observational study of aSAH cases and matched controls from the UK Biobank. The presence of fatigue was compared between cases and controls using the chi‐squared test. The change in frequency over time was assessed using Spearman's rank correlation coefficient. The effect of fatigue on employment was assessed using mediation analysis. Results Fatigue is more common following aSAH compared to matched controls (aSAH 18.7%; controls 13.7%; χ 2 = 13.0, p < 0.001) at a mean follow‐up of 123 months. Fatigue gradually improves over time with significant fatigue decreasing by 50% from ~20% in the first year to ~10% after a decade (p = 0.04). Fatigue significantly mediated 24.0% of the effect of aSAH status on employment. Conclusions Fatigue is common following aSAH and persists in the long term. It gradually improves over time but has a major impact on aSAH survivors, significantly contributing to unemployment following haemorrhage. Further work is required to develop treatments and management strategies for fatigue with a view to improving this symptom and consequently employment following aSAH.


Fatigue
Fatigue was assessed in the UK Biobank at assessment centre visits using the question 'Over the past 2 weeks, how often have you felt tired or had little energy?' (data field 2080). Individuals were categorized as suffering significant fatigue if they reported tiredness or little energy for more than half the time. A subset of individuals answered questions about fatigue phenotype (Table 1), scored using a 7-point scale with a score of 1 indicating strong disagreement and 7 strong agreement. Where applicable correction for multiple testing was performed using the Benjamini-Hochberg procedure with a false discovery rate of 5%.

Aneurysmal subarachnoid haemorrhage population
Aneurysmal subarachnoid haemorrhage cases were identified from the UK Biobank using International Classification of Diseases (ICD) 9 (data field 41271), ICD-10 (data field 41270), self-reported medical conditions (data field 20002) and primary care data (data field 42040). Individuals were excluded if the subarachnoid haemorrhage was non-aneurysmal in nature or if there was a trauma code documented within 30 days of diagnosis (see Table S1 for inclusion and exclusion codes). aSAH cases were included in this study if they had data on fatigue subsequent to the diagnosis of aSAH.

Control population
A single matched control population was identified from the UK Biobank using propensity score matching with a nearest neighbour method and a case:control ratio of 1:4. Individuals were matched according to age at follow-up, sex, smoking status and presence of anxiety or depression which have been shown to influence fatigue following aSAH [8,10,11]. Smoking status was dichotomized into current smoker or not (data field 20116). Anxiety and depression were dichotomized on whether the individual had seen a doctor for nerves, anxiety, tension or depression (data field 2090). Individuals with missing data on fatigue or covariates were excluded from the control pool available for matching.

Primary analysis
The chi-square test was used to compare frequency of fatigue between cases and controls. The t test was used to compare fatigue phenotype domains. Spearman's rank correlation coefficient was used to assess the relationship between frequency of fatigue and time.
Severity of clinical presentation and complications of aSAH, such as hydrocephalus, have been shown to be predictive of fatigue [9,10]. Logistic regression was used to explore whether these features were associated with significant fatigue in this dataset. The dependent variable was significant fatigue with the variable of interest as the independent variable in addition to age, sex, smoking status and presence of anxiety/depression. The presence of hydrocephalus was defined using the Office of Population Censuses and Surveys  (WFNS) grade is a measure of the severity of clinical presentation and the strongest known predictor of outcome following aSAH [14].
WFNS grade is not available in the UK Biobank but length of stay, which is strongly correlated with WFNS [15], was used as a surrogate.

Mediation analysis
To explore whether significant fatigue mediated any component of the effect of aSAH on employment status, causal mediation analysis using a natural effects model was performed utilizing the package medflex [16]. This method has been shown to be superior when analysing a binary mediator and outcome [17]. A non-parametric bootstrap procedure with 1000 samples was used to derive standard errors and p values. This was performed in the aSAH and matched control cohorts, additionally controlling for the Townsend deprivation score [18] (data field 189) and education status, dichotomized into people holding a college or university degree at the time of initial assessment in the UK Biobank or not (data field 6138). Employment status was dichotomized into good and poor, with poor employment defined as 'unemployed' or 'unable to work because of sickness or disability' (data field 6142). The proportion of the effect of aSAH status on employment mediated by fatigue was calculated using the method described by VanderWeele [19].
To provide context and assess the relative importance of fatigue to employment a further mediation analysis was performed exploring what proportion of the effect of aSAH status on employment was mediated by persistent headache, another common sequela of aSAH [20]. Headache was defined as present or absent using data field 6159 ('In the last month have you experienced headache that interfered with usual activity?') and the same causal mediation analysis was performed.
All analyses were performed in R (version 3.6.2, R Foundation for Statistical Computing).

RE SULTS
A total of 869 aSAH cases were identified from the UK Biobank. 829 were eligible for inclusion with data available on fatigue. 479,617 individuals were eligible for inclusion in the control cohort and 3316 controls were matched with a mean standard difference <0.004 (see Table 2 for demographics of individuals included in the study and Figure 1 for the flowchart of aSAH cases included).

Primary analysis
Significant fatigue was more frequent in cases compared to controls  (Table 3). This suggests that fatigue has an impact in almost all domains of life and significantly impairs a patient's quality of life.
The frequency of significant fatigue decreased by half from 19.6% in the first year following aSAH to 11.1% in the eleventh year with a significant relationship between frequency of fatigue and time (R S = −0.62, p = 0.04, Figure 2).

Mediation analysis
Unemployment or inability to work due to sickness/disability was significantly more frequent in the aSAH population (aSAH 18.7%; con-

DISCUSS ION
In this large sample size, it is demonstrated that fatigue is more common following aSAH compared to matched controls and persists in the long term, with a mean follow-up of over 10 years. Significant fatigue, defined as present for greater than 50% of the time, gradually improves over time in about half of patients, but has important implications. aSAH survivors report that it is one of the most disabling symptoms impacting quality of work, social and family life. In keeping with this, it is demonstrated that fatigue makes a large contribution to unemployment and inability to work due to sickness/ disability following aSAH. This information will be helpful to counsel patients regarding the duration and prognosis of fatigue following aSAH and emphasizes the importance of management strategies to improve this disabling symptom and consequently promote a return to employment.
Kutlubaev et al. [7] reported a weighted mean frequency of fatigue of 73.6% in the first year following aSAH using five studies. patients also reported that the prevalence of fatigue gradually decreased from 1 to 7 years post-aSAH, although the decrease was not statistically significant [10]. In this study a larger sample size is included and a longer follow-up explaining the greater significance of our results. It is also shown that length of stay, a surrogate of severity of clinical presentation, and hydrocephalus are not predictors of fatigue following aSAH. These results differ from the same recent study [10]. This may be because the UK Biobank favours good outcome individuals due to the requirement to engage in detailed follow-up assessments. Both severity of clinical presentation and hydrocephalus are predictive of poor outcome [23] and are consequently underrepresented in this cohort, limiting our ability to study their association with fatigue. However, it may also be a real observation. It would be easy to rationalize that, once treated, hydrocephalus does not increase fatigue, supported by a further study which showed an association between acute but not chronic treatment of hydrocephalus [9]. Also, although it would be easy to assume more severe haemorrhages result in more severe fatigue, it is possible that patients with worse outcomes have lower activity levels and are more focused on their functional deficits and relatively underreport fatigue. This fits with our anecdotal observations that often some of the best performing patients are most limited by fatigue.
Unemployment is common following aSAH with up to 50% reporting impaired return to work [24]. A number of factors have been implicated in return to work following aSAH including independence at discharge, consciousness at admission [25] and cognitive deficits following aSAH [2]. Fatigue has also been implicated [8,9] and this study emphasizes the importance of fatigue to employment by demonstrating that it mediates a significant propor- with cognition also contributing a much smaller effect on employment (24.0% vs. 6.6% [2]).
Both fatigue and unemployment impair quality of life following aSAH [8,26] emphasizing the importance of managing the symptom of fatigue following aSAH, especially as it persists in the long term and impacts employment. At present there are no pharmacological therapies to improve fatigue following stroke [27], but there are non-pharmacological strategies which can improve the symptoms of fatigue [28]. Uptake of these strategies following aSAH in addition to ongoing pharmacological trials (e.g., NCT 03209830) may help to improve fatigue with subsequent benefits for survivors' employment and quality of life.

Limitations
As UK Biobank participants are required to attend multiple very detailed assessment centre visits this study is biased towards individuals with a better outcome and more motivation. In comparison to poor outcome individuals who are preoccupied by functional deficits, aSAH cases included in this study are more likely to be aware of symptoms such as fatigue. Consequently, caution should be taken when applying these results to poorer outcome individuals.
In this study, a single question ('Over the past 2 weeks, how often have you felt tired or had little energy?') was used to assess

TA B L E 3
Comparison of fatigue phenotype questions between aneurysmal subarachnoid haemorrhage (aSAH) and control cohorts using the t test F I G U R E 2 Change in frequency of fatigue over time, divided into 12-month bins. Data beyond 11 years were not included due to the sparsity of data in each annual bin